The Rise of Mixture of Experts: How Sparse Architectures Are Revolutionizing Large Language Model Performance